Geo log i cal Quar terly, 2018, 62 (4): 755–766 DOI: http://dx.doi.org/10.7306/gq.1438
An a logue GPR study of the Perm ian fanglomerates from Zygmuntówka Quarry near Chêciny, Holy Cross Moun tains, south ern Po land, for con struc tion
of a train ing im age for mul ti ple point sim u la tions
Tomasz ¯UK1, *, Zbigniew MA£OLEPSZY1 and Ewa SZYNKARUK1
1 Pol ish Geo log i cal In sti tute – Na tional Re search In sti tute, Rakowiecka 4, 00-975 Warszawa, Po land
¯uk, T., Ma³olepszy, Z., Szynkaruk, E., 2018. An a logue GPR study of the Perm ian fanglomerates for con struc tion of a train - ing im age for mul ti ple point sim u la tions, Zygmuntówka Quarry, Holy Cross Moun tains, south ern Po land. Geo log i cal Quar - terly, 62 (4): 755–766, doi: 10.7306/gq.1438
As so ci ate ed i tor: Anna Wysocka
The dis tri bu tion of Perm ian al lu vial fan lithofacies in a quarry at Zygmuntówka near Chêciny, Holy Cross Mts., in south ern Po land was in ves ti gated us ing ground pen e trat ing ra dar (GPR) in or der to cre ate a train ing im age for mul ti ple point sta tis tics (MPS) re con struc tions of al lu vial fan sed i men tary fa cies. Five pseudo-3D GPR datasets were col lected, pro cessed and up - loaded for in ter pre ta tion into SKUA-GOCAD 3D geo log i cal mod el ling soft ware. Three ra dar fa cies were dis tin guished based on the 3D geo met ri cal pat tern of ra dar re flec tions and linked to lithofacies de scribed from the quarry by Zbroja et al. (1998). A sta tis ti cal sum mary showed that ~50% of the lithofacies re sulted from grav ity flows (mostly non-co he sive), while the re main - ing pro por tion was de pos ited by un con fined and con fined flash floods. Flu vial sed i men tary fa cies left by wan ing of cat a - strophic floods or re work ing dur ing fair weather, al though not prev a lent, could not be dis tin guished from con fined flood de pos its based only on GPR data. The GPR datasets to gether with in for ma tion from field ob ser va tions were used to carry out MPS sim u la tions and es ti mate the most prob a ble 3D model of lithofacies at the quarry scale. This model will in turn serve as a train ing im age for MPS re con struc tions of al lu vial-fan fa cies of Rotliegend con glom er ates in the multi-scale geo log i cal model of the Gorzów Block (west ern Po land).
Key words: al lu vial fans, ground pen e trat ing ra dar, Perm ian con glom er ates, sed i men tary fa cies, mul ti ple point sta tis tics, train ing im age.
INTRODUCTION
Multi-scale 3D geo log i cal mod el ling of the sed i men tary cover of the Gorzów Block (west ern Po land; Fig. 1), apart from de lin eat ing the strati graphic sys tems and main tec tonic struc - tures, also in volves re con struct ing the spa tial dis tri bu tion of sed i men tary fa cies for the most im por tant ho ri zons over ly ing the Variscan base ment. In the Gorzów Block, as in other post-Variscan intra-con ti nen tal bas ins of West ern and Cen tral Eu rope, the tec toni cally de formed base ment of Pre cam brian to Car bon if er ous age is dis cor dantly cov ered by the Perm ian and youn ger sys tems. How ever, the great depth of these strata and sparse bore hole data mo ti vate us to use an a logues of the dom i - nant lithostratigraphic units to prop erly de lin eate fa cies dis tri bu - tion. A sig nif i cant pro por tion of the Perm ian sed i men tary cover on the Gorzów Block is formed of al lu vial-fan con glom er ates which ac cu mu lated around vol ca nic el e va tions on the north -
west ern part of the Wolsztyn High (Fig. 1). These fa cies – ac - ces si ble only in deep bore holes in the Gorzów Block – were de - scribed by Pokorski (1988) as al lu vial fan con glom er ates of the basal part of the Noteæ For ma tion. Karnkowski (1994) de - scribed them as an allostratigraphic unit of the Ksi¹¿
Wielkopolski Con glom er ate, while Kiersnowski and Buniak (2006) as con glom er ates of the ALIII depositional cy cle. The near est out crops of com pa ra ble Perm ian fa cies can be found in the Sudetes, near Kraków and near Chêciny in the Holy Cross Mts. (Fig. 1). Quan ti ta tive re la tion ships mea sured in an a logue geo log i cal sit u a tions com pared with fa cies mod els es tab lished based on ob ser va tions from con tem po rary al lu vial fans (for ex - am ple Nemec and Steel, 1984; Blair and McPherson, 1992, 1994a, b) can aid re con struc tions of sed i men tary fa cies dis tri - bu tion in in ac ces si ble sit u a tions, i.e. be tween deep bore holes.
One ap proach to re con struct ing this fa cies in volves us ing mul ti - ple point sta tis tics (MPS), which re quires train ing im ages.
Train ing im ages are con cep tual mod els rep re sent ing the spa tial ar range ment of prop er ties. They are con structed based on ob ser va tions from ex po sures, high-res o lu tion geo phys i cal data such as ground pen e trat ing ra dar (GPR), un der stand ing of fa cies mod els and sed i men tary pro cesses, 3D geo met ri cal mod el ling tech niques or other quan ti ta tive meth ods (Comunian et al., 2011; Linde et al., 2015). Train ing im ages of fer an al ter -
* Corresponding author, e-mail: tomasz.zuk@pgi.gov.pl
Re ceived: March 14, 2016; ac cepted: May 15, 2018; first pub lished on line: November 23, 2018
na tive to variogram-based geostatistics, which deals with spa - tial re la tion ships be tween two mea sure ment points, but can not re solve com plex pat terns such as the ar range ment of sed i men - tary fa cies (Guardiano and Srivastava, 1992; Strebelle, 2002).
Dur ing sim u la tions, the MPS al go rithm scans the train ing im - age, ex tracts re la tion ships be tween mul ti ple data points and ap plies them to avail able subsurface data from a new site in or - der to stochastically rec re ate mul ti ple, equally pos si ble, new dis tri bu tions.
Nu mer ous stud ies of al lu vial fanglomerates show a great sim i lar ity of al lu vial fan ar chi tec ture, par tic u larly their main sed i - men tary fa cies (cf. Nemec and Steel, 1984; Nemec and Postma, 1993; Blair and McPherson, 1994a, b; Maejima and Nakanishi, 1994; Blair, 2000; Kim and Lowe, 2004). This is due to the dom i nant role of grav ity flow pro cesses as well as cat a - strophic floods both in desert and non-desert en vi ron ments.
Typ i cal al lu vial fan fa cies are pri mar ily formed by sub-ae rial grav ity flows and cat a strophic (flash) floods (super criti cal flow of New to nian fluid) as de scribed by Nemec and Steel (1984) and Blair and McPherson (1994b). Flu vial pro cesses as so ci ated with subcritical flow oc cur on al lu vial fans only dur ing the wan - ing phase of cat a strophic flows (form ing the top part of the se - quence) or in fair-weather con di tions when they only mod ify the al lu vial fan sur face. Most sur face ob ser va tions from al lu vial fans were car ried out dur ing dry weather which led to the com - mon as so ci a tion of al lu vial fan en vi ron ments with flu vial pro - cesses (Blair and McPherson, 1994a).
Bed rock li thol ogy and cli mate are two pri mary con trol ling fac tors for weath er ing pro cesses, which af ter de po si tion of sed -
i men tary ma te rial may re sult in dif fer ences in ma trix amount and type (Thiry et al., 1999; Tay lor and Eggleton, 2001). How - ever, a greater amount of fine par ti cles will only fa vour co he sive de bris flows over non-co he sive flows dur ing the ini tial stage of bed rock ero sion and fan de vel op ment (type I al lu vial fans of Blair and McPherson 1994a). Dur ing MPS sim u la tions, pro por - tions of sed i men tary fa cies can be im posed based on hard data such as bore hole in for ma tion. Thus for the pur pose of cre at ing a train ing im age, it is pos si ble to com pare al lu vial fans built by sed i men tary ma te rial of dif fer ent com po si tions, such as fans com pris ing car bon ate clasts as is the case at Zygmuntówka Quarry, with those com posed of frag ments of vol ca nic or clastic sed i men tary rocks (Gorzów Block).
The aim of this ar ti cle is to dem on strate the pro cess of de vel - op ing a train ing im age based on pseudo-3D ground pen e trat ing ra dar (GPR) data col lected from out crops and cor re lated with ex - po sure ob ser va tions. We dis cuss the ap pli ca bil ity of such an ap - proach us ing data col lected from the Zygmuntówka Quarry near Chêciny in south ern Po land. The phase of ap ply ing the train ing im age in re con struct ing sed i men tary lithofacies with MPS based on the ex am ple of the Gorzów Block is out side the scope of this ar ti cle and will be de scribed in a sep a rate pub li ca tion.
GEOLOGICAL SETTING
Coarse-grained clast-sup ported, oc ca sion ally ma trix-sup - ported, con glom er ates of Late Perm ian age are ex posed in the Zygmuntówka Quarry at the hill of Czerwona Góra, lo cated near Fig. 1. Lo ca tion of the study site near Chêciny in south ern Po land and the ex tent
of the Gorzów Block model
Black dashed line in di cates lo ca tion of the Wolsztyn High
the town of Chêciny (Holy Cross Mts.) ~10 km south of Kielce, south ern Po land (Figs. 1 and 2). The Perm ian strata have not been sig nif i cantly tec toni cally de formed here, bed ding be ing gently in clined at an an gle of ~5 to the south-west, i.e. the dom i - nant trans port di rec tion (Kostecka, 1962). Thus the study site of fers ex cel lent con di tions to study the depositional ar chi tec ture of an an cient al lu vial fan.
The al lu vial-fan fa cies un con form ably over lie de formed Up - per De vo nian and lo cally Lower Car bon if er ous car bon ates form ing palaeopiedmont set tings for the Perm ian de pos its of an arid and semi-arid cli mate. Here, the thick ness of the Perm ian suc ces sion reaches over 150 m as re corded in bore hole BZ1 lo cated ~1.5 km to the south-east of the study site (Kuleta et al., 2007). Only the up per most 30 m thick sec tion of the Up per Perm ian con glom er ates is ex posed in the quarry and is in ter - preted to have been de pos ited as a prox i mal to in ter me di ate al - lu vial fan. Gen eral up wards coars en ing of clasts sug gests a prograding fan sys tem (Zbroja et al., 1998). The lower part of the pro file lo cated to the south-east of Zygmuntówka was in ter - preted to con sist of shal low fan-delta de pos its, as sug gested by the pres ence of lime stone beds and the chem i cal com po si tion of the ma trix (Migaszewski et al., 1996; Kuleta et al., 2007).
Zbroja et al. (1998) de scribed five lithofacies within the Zygmuntówka con glom er ates (Fig. 2). The first two are thickly-bed ded, structureless, clast-sup ported (lithofacies Z1) and ma trix-sup ported (lithofacies Z2), poorly sorted con glom er - ates with cob bles and boul ders. Lithofacies Z1 form ing ~40% of the rock vol ume was de pos ited by non-co he sive grav ity flows (Gani, 2004), also de fined as pseudoplastic grain flows (Nemec, 2009) or the liq ue fied flows of Middle ton and Hampton (1973). Ma trix-sup ported grav els with cob bles and boul ders (Z2) are co he sive grav ity flow de pos its (Bingham plas tic rhe ol - ogy, Nemec 2009) and rep re sent only 10% of the rock vol ume.
The third lithofacies is de scribed as subhorizontally bed ded, clast-sup ported, poorly sorted con glom er ate (lithofacies Z3) with clast di men sions reach ing 0.3 m, com monly nor mally graded. Lithofacies Z3 is in ter preted as de pos its as so ci ated with un con fined, super criti cal flow pro duced by cat a strophic flash floods (sheetfloods). Trough cross-bed ded con glom er - ates with oc ca sional nor mal grad ing were clas si fied as lithofacies Z4 as cribed to de po si tion from small streams.
Structureless clast-sup ported con glom er ates are ar ranged in both pla nar and trough-shaped beds and de scribed as lithofacies Z5. They are as cribed to con fined (channelized) An a logue GPR study of the Perm ian fanglomerates from Zygmuntówka Quarry near Chêciny... 757
Fig. 2. Field sketch of the quarry with lo ca tion of GPR datasets and the geo log i cal pro file from Zbroja et al. (1998) show ing dis tri bu tion of the main lithofacies Z1–Z5
Z1 – thickly-bed ded, structureless, clast-sup ported; Z2 – thickly-bed ded, structureless, ma trix-sup ported; Z3 – thinly- to me dium-bed ded, clast-sup ported, poorly sorted; Z4 – trough cross-bed ded con glom er ates, oc ca sion ally with nor mal grad ing; Z5 – structureless, clast-sup - ported con glom er ates ar ranged in both pla nar and trough-shaped beds; symbols show ing clast size: cl – clay, si – silt, s – sand, fg – fine gravel, cg – coarse gravel; A–E – GPR datasets; num bers near GPR datasets in di ca tion their di men sions in metres; the ar row in di cates the north
flashfloods (as de scribed by Simons and Rich ard son, 1966) rep re sent ing in cised-chan nel fills and the em bank ments sur - round ing them (Fig. 3).
METHODS
GPR sends short elec tro mag netic im pulses which are re - flected back to the sur face from in ter faces be tween zones of con trast ing di elec tric prop er ties and re corded by a re ceiver. We used RAMAC GPR with a 250 MHz shielded an tenna. Five pseudo-3D GPR datasets were col lected with their lat eral di - men sions rang ing from 5 to 25 m (Fig. 2). Sig nal pen e tra tion reached on av er age a depth of 10 m. Each dataset com prised a se ries of par al lel 2D GPR pro files with a dis tance be tween lines of 0.2 m and a sur vey step of 0.1 m form ing a sur vey grid close to full-3D res o lu tion (Grasmueck et al., 2005; Dogan et al., 2011). Closely spaced par al lel GPR pro files were in di vid u ally pro cessed (fre quency bandpass fil ter ing, gains, in some datasets top o graphic cor rec tion), and 2D lines were com piled into 3D datasets. Large amounts of cob bles and boul ders in tro - duced hy per bola-shaped re flec tions which ob scured the im - ages, and only af ter the data was mi grated (2D Stolt) pro files were in ter pret able. Sig nal ve loc ity for the data mi gra tion, top o - graphic cor rec tion and time-to-depth con ver sion was es tab - lished at two lo ca tions in the quarry (at the cen tre of grids A and D) to be 0.087 m/ns based on the thick ness of strata known from the quarry ex po sures. The GPR method is anal o gous to seis mic re flec tion, there fore pro cessed pseudo-3D GPR data could be up loaded for in ter pre ta tion into SKUA-GOCAD geo - log i cal mod el ling soft ware. Field ob ser va tions were also car ried out in or der to ver ify the lo ca tion of sed i men tary fa cies in the quarry cliffs, par tic u larly in the top most quarry face, which was not avail able for ac qui si tion of GPR data due to veg e ta tion and an un even ground sur face.
RESULTS AND INTERPRETATION
CLASSIFICATION OF RADAR FACIES
Clas si fi ca tion of ra dar fa cies was based on con cepts of ra - dar stra tig ra phy (Beres and Heani, 1991; Gawthorpe et al.,
1993; Huggenberger, 1993) which were adopted from seis mic stra tig ra phy (Mitchum et al., 1977). Ra dar fa cies were de fined by Huggenberger (1993) as mappable, three-di men sional sed i - men tary units char ac ter ized by re flec tions the spa tial pat tern of which dis tin guishes it from ad ja cent units. The 3D ap proach to ra dar fa cies rep re sen ta tions in volves re cog nis ing a re flec tion pat tern in all three di rec tions (x-y-z). Large-scale sed i men tary struc tures (decimetre-scale) ap par ent on GPR pro files and in - ter faces be tween zones of con trast ing elec tro mag netic prop er - ties were man u ally picked across datasets. Hor i zon tal sec tions (time slices) were par tic u larly use ful be cause they al lowed the trac ing of ho ri zons be tween ad ja cent 2D lines within 3D datasets, which oth er wise would be dif fi cult to cor re late in some ar eas (Fig. 4). As in the case of seis mic data, the main prop er - ties which de scribe the char ac ter of re flec tion sur faces are their am pli tude, shape, dip, az i muth of in cli na tion, con ti nu ity as well as spa tial re la tion ships with other in ter faces, par tic u larly of higher or der (sur faces be tween ra dar fa cies).
Dip and az i muths of sur faces (man u ally traced) were au to - mat i cally cal cu lated by the mod el ling soft ware, dis played on sur faces and sum ma rized for in ter pre ta tion us ing his to grams.
Gen er ally, the steeply in clined, typ i cally at an an gle of 6 to 15 and lo cally reach ing 20 (Figs. 4B and 5), tab u lar ar range ment of ra dar re flec tions was clas si fied as ra dar fa cies Z1/Z2. This de scrip tion which merged fa cies Z1 and Z2 was made in or der to link the ra dar fa cies to the clas si fi ca tion of lithofacies of Zbroja et al. (1998) for the study site (ex plained in the in ter pre - ta tion sec tion). Subhorizontal con tin u ous par al lel re flec tions in - clined at an an gle of be tween 2 and 8 were grouped as ra dar fa - cies Z3. Trough-shaped con cor dant re flec tions, typ i cally semi - con tinu ous, were named ra dar fa cies Z4/Z5 (Fig. 5). These troughs were very ir reg u lar in shape, depth and lat eral ex tent, but on av er age their depth was 1–3 m with lat eral ex tents of 3–8 m. The sum mary of ra dar fa cies cor re lated with the lithofacies of Zbroja et al. (1998) is shown in Fig ure 6.
The most dis tinct (high est am pli tude) re flec tions were man - u ally traced and trans formed into sur faces us ing geomodelling soft ware (Figs. 5 and 6). Most re flec tions were dif fi cult to trace, par tic u larly the trough-shaped ones, as they con tinue lat er ally only for 2 to 5 m. This is con sis tent with de scrip tions of bed ding con ti nu ity from con tem po rary al lu vial fans (e.g., Blair, 2000).
Thus trac ing sur faces be tween 3D GPR datasets (cubes) was not re al is tic. The sur faces were clas si fied ac cord ing to the ra dar fa cies they be longed to; their in cli na tion and az i muth was ex - tracted us ing the soft ware, and sum ma rized as his to grams and ba sic sta tis ti cal val ues. Based on this sum mary, re flec tion sur - faces of ra dar fa cies Z1/Z2 were the most steeply in clined at an av er age an gle of 11° with a stan dard de vi a tion value of 6.3. In - cli na tion of sur faces clearly in creased with depth with max i mum val ues of over 20° in the bot tom parts of the quarry. Ra dar fa - cies Z3 was as so ci ated with an av er age dip of 7.6° and stan - dard de vi a tion of ~4. Some sur faces in the up per sec tions as so - ci ated with fa cies Z3 are very shal low and their an gle of in cli na - tion was <3°.
The av er age az i muth val ues for each of the ra dar fa cies are quite sim i lar, al though in the case of lithofacies Z3 the mean az i - muth of in cli na tion was 232.8°, the most sim i lar to the az i muth of in cli na tion of the De vo nian subcrop (234°). This subcrop sur - face was traced based on the De tailed Geo log i cal Map of Po - land at the scale of 1:50,000, Chêciny Sheet (Hakenberg, 1971). For this fa cies, the stan dard de vi a tion is the high est (~53°) and the dis tri bu tion is clearly bi modal which may re flect the con vex shape of the al lu vial fan to pog ra phy. The az i muth for ra dar fa cies Z1/Z2 is the most west erly (245°), while the stan dard de vi a tion is the low est (26°).
Fig. 3. Lithofacies Z5 in the up per most quarry face, photo fac ing the north
An a logue GPR study of the Perm ian fanglomerates from Zygmuntówka Quarry near Chêciny... 759
Fig. 4. GPR dataset A
A – with di men sions 12 ´ 7 m lo cated in the east ern part of the low est bench of the quarry; note the az i muth of in cli na tion also can be in ferred from the hor i zon tal sec tion (at the top of the dataset); these hor i zon tal sec tions helped to trace the ho ri zons be tween par al lel 2D GPR pro files form ing each of the pseudo-3D datasets; to tal ver ti cal reach in two-way-travel time (TWTT) is 220 ns (100 ns = 4.35 m); B – with some of the ho ri zons dis played and show ing dip in cli na tion (in de grees); note the very high an gles of in cli na tion on the sec ond, third and bot tom ho ri zons (lo cally reach ing 20 ); the sec ond ho ri zon from the bot tom with val ues be low 10 is as so ci ated with ra dar fa cies Z3; the un du la tions on the sur - faces may have re sulted from data ac qui si tion, pro cess ing and in ter pre ta tion, i.e. man ual pick ing of sur faces; total ver ti cal reach in TWTT is 220 ns (100 ns = 4.35 m)
SEDIMENTARY INTERPRETATION OF RADAR FACIES
Link ing the ra dar fa cies in GPR datasets with the lithofacies de scribed in the quarry ex po sures and there fore also the pro - cesses that shaped them was pos si ble us ing the main at trib - utes de scrib ing ra dar fa cies: shape of re flec tion sur faces (in - clud ing their con ti nu ity), their am pli tude, dip and az i muth of in - cli na tion. As iden ti fied by Blair and McPherson (1994b), bed - ding in cli na tion within al lu vial fan de pos its can sug gest the depositional pro cesses re spon si ble for their for ma tion. Thus, a par al lel steeply in clined (8 to 20°) tab u lar ar range ment of ra dar re flec tions, oc ca sion ally with deeply in cised scour-shaped re - flec tions, as in the case of dataset D, can cer tainly be in ter - preted as lithofacies Z1 or Z2 de scribed by Zbroja et al. (1998).
These sed i ments were de pos ited by co he sive and non - cohesive de bris flows, rep re sented re spec tively by ma trix- and frame work-sup ported grav els. They could not be dis tin guished be tween each other based only on the pat tern of GPR re flec - tions, while ex po sure cor re la tion was not al ways avail able. A sta tis ti cal sum mary done with the geomodelling soft ware
showed that fa cies Z1 and Z2 to gether rep re sent about 50% of lithofacies in all GPR datasets, which sup ports ob ser va tions in quarry ex po sures made by Zbroja et al. (1998).
The subhorizontal par al lel strat i fi ca tion sug gested litho - facies Z3 was de pos ited by lam i nar cat a strophic flood flows.
The in cli na tion of this strat i fi ca tion was <8°. Al though chan nels can some times also be pres ent within de bris flow de pos its, chan nels in al lu vial fans are most of ten as so ci ated with con - fined flash flood de pos its (Blair and McPherson, 1994a). Thus trough-shaped con cor dant re flec tions were used to iden tify lithofacies Z4 or Z5. Lithofacies Z4 was de pos ited by wan ing trac tional cur rents as de scribed by Nemec and Steel (1984) or per haps are de bris flow or flash flood de pos its re worked dur ing pe ri ods of fair weather (Z4). Al though re flec tions re sem bling cross-strat i fi ca tion were no ticed within chan nels on some of the pro files, they could not be clearly dis tin guished from lithofacies Z5, i.e. channelized flashflood de pos its. There fore ra dar fa cies Z4 and Z5 have been merged. Ra dar fa cies cor re lated with the lithofacies of Zbroja et al. (1998) are sum ma rized in Fig ure 6.
Fig. 5. GPR dataset B (lo cated on the up per bench) with ho ri zon in cli na tion (dip in de grees) dis played on some of the ho ri zons
The sec ond ho ri zon (trough-shaped) as so ci ated with fa cies Z4/Z5 does not show dip for clar ity of the im age;
to tal ver ti cal reach in TWTT is 220 ns (100 ns = 4.35 m)
MULTIPLE POINT STATISTICS (MPS)
MPS sim u la tions were car ried out us ing the Im proved Par - al lel Mul ti ple Point Al go rithm (IMPALA, Straubhaar et al., 2010) by Ephesia in cor po rated in SKUA-GOCAD soft ware. In con - trast to two-point geostatistics (for ex am ple krigining), which is based on semivariogram and re la tion ships for pairs of mea - sure ment points, mul ti ple point sta tis tics takes into ac count spa tial re la tion ships be tween mul ti ple mea sure ment points (Guardiano and Srivastava, 1992). These re la tion ships be - tween data points can be ob tained from train ing im ages (Strebelle, 2000, 2002) in or der to re pro duce sim i lar geo met ri - cal pat terns of value dis tri bu tion. MPSs have been ap plied to pre dict the dis tri bu tion of prop er ties in in ac ces si ble geo log i cal sit u a tions, for ex am ple the dis tri bu tion of hy drau lic con duc tiv ity within hy dro car bon res er voir rocks or aqui fers (Ronayne et al., 2008; Comunian et al., 2011; Linde et al., 2015).
The 3D ra dar fa cies were painted on each of pseudo-3D GPR datasets (see dataset C in Figs. 7 and 8) in SKUA-GOCAD. Field ob ser va tions of the po si tion of lithofacies Z5 (up per most quarry face) and Z1 (low er most cliff), to gether with ra dar fa cies from GPR datasets, were trans ferred to a sin - gle geo log i cal grid with lat eral di men sions of 120 ´ 80 m and a height of 35 m. The grid was built of cells, each with di men sions of 1.2 ´ 1.2 ´ 0.45 m. This grid was used to sim u late the dis tri - bu tion of lithofacies at the quarry scale. Thirty re al iza tions from
MPS sim u la tions were cre ated us ing all avail able in for ma tion on the lithofacies dis tri bu tion from five GPR datasets and field ob ser va tions. GPR dataset C (Fig. 8) was used as a train ing im - age be cause its fa cies pro por tions were the most sim i lar to the pro por tion of fa cies es tab lished us ing all GPR datasets.
The re al iza tions from MPS sim u la tions were used to com - pute prob a bil i ties us ing SKUA-GOCAD to find the most likely so lu tion. Quantile val ues be tween 0 and 1 rep re sent con tig u - ous ranges of equal prob a bil ity. In the case of 10-quantile, 0.5 quantile is the same as the me dian on a cu mu la tive prob a bil ity curve, i.e. the mid dle value on a cu mu la tive prob a bil ity curve.
This re sult is used as the most prob a ble (Fig. 9A). In or der to graph i cally dem on strate un cer tainty as so ci ated with this es ti - ma tion, reali sa tions for quantiles be tween 0.2 and 0.8 are shown in Fig ure 9B–E.
DISCUSSION
The ra dar fa cies were clas si fied and in ter preted as sed i - men tary fa cies mostly based on shape of re flec tions and their dip of in cli na tion. These fea tures can not be in ferred from 2D data (2D GPR pro files or rock ex po sures), par tic u larly in such com plex depositional sys tems as al lu vial fans. The study dem - on strated how 3D geo phys i cal meth ods sup ported by sto chas - An a logue GPR study of the Perm ian fanglomerates from Zygmuntówka Quarry near Chêciny... 761
Fig. 6. Clas si fi ca tion of ra dar fa cies to gether with cor re la tion with lithofacies and sed i men tary in ter pre ta tion, units in the 3D rep re sen ta tions ex pressed in metres
Fig. 7. 2D oblique sec tion through dataset C or thogo nally to the dip di rec tion of the up per most sur face Only bound ing sur faces are dis played for the clar ity of the im age; com pare with Figure 8; lat eral di men sions in m, ver ti cal scale
in show ing TWTT in ns (100 ns = 4.35 m)
Fig. 8. GPR dataset C lo cated in the west ern part of the sec ond-from-the-bot tom level of the quarry with lithofacies Z1/2 (grey), Z3 (yel low) and Z4/5 (or ange)
Lat eral di men sions in metres, ver ti cal scale in show ing TWTT in hun dreds of ns (100 ns = 4.35 m)
An a logue GPR study of the Perm ian fanglomerates from Zygmuntówka Quarry near Chêciny... 763
Fig. 9. Re sult of MPS es ti ma tion for quantile
A – 0.5 based on 30 sim u la tions of fa cies dis tri bu tion in the quarry-wide geo log i cal grid with lat eral di men sions of 120 ´ 80 m and a height of 35 m; B – 0.2; C – 0.4 to dem on strate un cer tainty as so ci ated with MPS sim u la tions based on pseudo-3D GPR data; ra dar fa cies: Z1/2 (grey), Z3 (yel low) and Z4/5 (or ange); grey sur face rep re sents the ground sur face around the Zygmuntówka Quarry
tic mod el ling can be used to re con struct sed i men tary fa cies dis - tri bu tion at a lo cal scale in or der to con struct re li able train ing im - ages for fur ther MPS sim u la tions of facies distribution in new geological situations.
The sed i men tary fa cies of fanglomerates iden ti fied with GPR com prise co he sive and non-co he sive grav ity flow de pos - its (Z1 and Z2) as well as fluid grav ity flow de pos its, both un con - fined sheetflood (Z3) and con fined chan nel fills (Z5). Zbroja et al. (1998) also iden ti fied lithofacies Z4 ei ther re lated to wan ing trac tion cur rents or re work ing of de bris flow or flash flood de - pos its dur ing pe ri ods of nor mal weather. Ac cord ing to Zbroja et al. (1998), this lithofacies was lim ited to two beds of lat eral ex - tent of 4 m and to tal thick ness reach ing 3.2 m. Be cause the last two fa cies are both as so ci ated with trough-shaped sur faces, they can not be dif fer en ti ated from lithofacies Z5 on ra dar im - ages. Thus, mod el ling of the dis tri bu tion of stream flow de pos its (lithofacies Z4) had to be aban doned and lithofacies Z5 was given a greater im por tance.
The sed i men tary ar chi tec ture sug gests that the study site pre serves a tran si tional sec tion of al lu vial fan be tween what Blair and McPherson (1994a) de scribed as a type I al lu vial fan (in the lower sec tion of the quarry pro file) and type II (above).
Type I al lu vial fans are dom i nated by de bris flows, mass-wast - ing pro cesses and pos si ble in cised chan nels and gul lies in their up per part, typ i cally with av er age slope in cli na tion be tween 5 and 15, while type II is dom i nated by un con fined and con fined flash floods with slope in cli na tion of be tween 2 and 8.
The train ing im age con structed from this study site us ing all data avail able – the most prob a ble fa cies dis tri bu tion at en tire quarry scale, as il lus trated in Fig ure 9A – shows the ar range - ment of all ma jor fa cies. This train ing im age should be suit able to carry out sim u la tions of sed i men tary fa cies dis tri bu tion in most palaeo geo graphi cal set tings and stages of fan de vel op - ment. The De vo nian subcrop sur face was dig i tally de lin eated based on the Geo log i cal Map of Po land at the scale of 1:50,000 (Hakenberg, 1971). This al lows re fer ring of the train ing im age to Fig. 9. Re sult of MPS es ti ma tion for quantile
D – 0.6; E – 0.8 to dem on strate un cer tainty as so ci ated with MPS sim u la tions based on pseudo-3D GPR data; ra dar fa cies: Z1/2 (grey), Z3 (yel low) and Z4/5 (or ange); grey sur face rep re sents the ground sur face around the Zygmuntówka Quarry
the lo cal paleaogeography and there fore its use in MPS sim u la - tions in dif fer ent palaeotectonic sit u a tions.
In the Gorzów Block, the sed i men tary ma te rial build ing al lu - vial fan de pos its mostly in cludes frag ments of Lower Perm ian vol ca nic rocks, al though in some ar eas (bore hole Banie 1) only ma te rial de rived from Car bon if er ous sed i men tary rocks was found within the con glom er ates. As ex plained in the in tro duc - tion, dur ing the phase of fa cies re con struc tion (MPS sim u la - tions) dif fer ences in the lithological com po si tion of sed i men tary ma te rial can be ad dressed by set ting fa cies pro por tions based on hard data from bore holes. The phase of ap pli ca tion of the training image will discussed in a separate publication.
CONCLUSIONS
The study high lighted ap pli ca tion of pseudo-3D GPR data to cre ate a train ing im age for MPS sim u la tions. The al lu vial fan
ar chi tec ture, dom i nated by de pos its of both de bris flows and flash floods, was in ves ti gated us ing pseudo-3D GPR data col - lected from the quarry ex po sures. It was sta tis ti cally ana lysed and lithofacies dis tri bu tion across the quarry was re con structed in or der to cre ate a train ing im age, i.e. the most prob a ble (quantile 0.5) dig i tal model of fa cies dis tri bu tion. This model can be used in anal o gous sit u a tions, and will be tested on al lu vial fan facies in the Permian system of the Gorzów Block.
Ac knowl edge ments. The study was car ried out as part of PGI-NRI pro ject: 3D Geo log i cal Model of Sed i men tary Cover of the Gorzów Block, pro ject No. 22.2609.1501.00.1. The au thors would like to thank the ed i tor A. Wysocka as well as R.L. Van Dam and an anon y mous re viewer whose con tri bu tion sig nif i - cantly im proved the manu script. We also would like to ac knowl - edge fruit ful dis cus sions with H. Kiersnowski from PGI-NRI.
Par a digm is thanked for mak ing SKUA-GOCAD avail able for this pro ject.
REFERENCES
Beres, M., Haeni, F.P., 1991. Ap pli ca tion of ground-pen e trat ing-ra - dar meth ods in hydrogeologic stud ies. Ground Wa ter, 29:
375–386.
Blair, T.C., 2000. Sedimentology and pro gres sive tec tonic un con - formi ties of the sheetflood-dom i nated Hell's Gate al lu vial fan, Death Val ley, Cal i for nia. Sed i men tary Ge ol ogy, 132: 233–262.
Blair, T.C., McPherson, J.G., 1992. The Trollheim al lu vial fan and fa cies model re vis ited. GSA Bul le tin, 104: 762–769.
Blair, T.C., McPherson, J.G., 1994a. Al lu vial fan pro cesses and forms. In: Geo mor phol ogy of Desert En vi ron ments (eds. A.D.
Abrahams and A.J. Par sons): 14–402. Chap man and Hall, Lon - don.
Blair, T.C., McPherson, J.G., 1994b. Al lu vial fans and their nat u ral dis tinc tion from rivers based on mor phol ogy, hy drau lic pro - cesses, sed i men tary pro cesses, and fa cies as sem blages. Jour - nal of Sed i men tary Re search, 64: 450–489.
Comunian, A., Renard, P., Straubhaar, J., Bayer, P., 2011.
Three-di men sional high res o lu tion fluvio-gla cial aqui fer an a log – Part 2: Geostatistical mod el ing. Jour nal of Hy drol ogy, 405:
10–23.
Dogan, M., Van Dam, R.L., Bohling, G.C., But ler, J.J., Hyndman, D.W., 2011. Hydrostratigraphic anal y sis of the MADE site: in te - gra tion of full res o lu tion GPR and hy drau lic con duc tiv ity data.
Geo phys i cal Re search Let ters, 38: L06405.
Gani, M.R., 2004. From tur bid to lu cid: A straight for ward ap proach to sed i ment grav ity flows and their de pos its. The Sed i men tary Re cord, 2: 4–8.
Gawthorpe, R.L., Col lier R.E.L., Al ex an der, J., Leeder, M., Bridge, J.S., 1993. Ground pen e trat ing ra dar, ap pli ca tion to sand body ge om e try and het er o ge ne ity stud ies. Geo log i cal So - ci ety of Lon don Spe cial Pub li ca tion, 73: 421–432.
Grasmueck, M., Weger, R., Horstmeyer, H., 2005. Full-res o lu tion 3D GPR im ag ing. Geo phys ics, 70: K12–K19.
Guardiano, F.B., Srivastava, R.M., 1992. Multivariate geostatistics: be yond bivariate mo ments. Quan ti ta tive Ge ol ogy and Geostatistics, 5: 133–144.
Hakenberg, M., 1971. Szczegó³owa mapa geologiczna Polski w skali 1:50 000, arkusz 850, Chêciny (in Pol ish). Instytut Geologiczny, Warszawa.
Huggenberger, P., 1993. Ra dar fa cies: rec og ni tion of fa cies pat - terns and heterogeneities within Pleis to cene Rhine grav els, NE Swit zer land. Geo log i cal So ci ety Spe cial Pub li ca tions, 75:
163–176.
Karnkowski, P.H., 1994. Rotliegend lithostratigraphy in the cen tral part of the Pol ish Perm ian Ba sin. Geo log i cal Quar terly, 38 (1):
27–42.
Kiersnowski, H., Buniak, A., 2006. Evo lu tion of the Rotliegend Ba - sin of north west ern Po land. Geo log i cal Quar terly, 50 (1):
119–138.
Kim, B.C., Lowe, D.R., 2004. Depositional pro cesses of the grav - elly de bris flow de pos its, South Do lo mite al lu vial fan, Owens Val ley, Cal i for nia. Geosciences Jour nal, 8: 153–170.
Kostecka, A., 1962. Char ac ter is tic of Zechstein con glom er ates of Ga³êzice–Bolechowice Syncline (Holy Cross Moun tains) (in Pol ish with Eng lish sum mary). Kwartalnik Geologiczny, 6 (3):
416–435.
Kuleta, M., Zbroja, S., Nawrocki, J., 2007. Perm ian con glom er ates of Zygmuntówka quarry. In: Fieldtrip guide for In ter na tional Work shop on the Tri as sic of South ern Po land, Fourth Meet ing for Pan-Eu ro pean Cor re la tion of the Epicontinental Tri as sic (eds. J. Szulc and A. Becker): 64–67. Pañstwowy Instytut Geologiczny, Warszawa.
Linde, N., Lochbühler, T., Dogan, M., Van Dam, R.L., 2015.
Tomogram-based com par i son of geostatistical mod els: ap pli ca - tion to the macrodispersion ex per i ment (MADE) site. Jour nal of Hy drol ogy, 531: 543–556.
Maejima, W., Nakanishi, T., 1994. Mid dle Mio cene al lu vial fan-fan delta sed i men ta tion: the Kanaso con glom er ate and sand stone mem ber of the Togane for ma tion to the North of Ha ma da, South - west Ja pan. Jour nal of Geosciences, Osaka City Uni ver sity, 37:
55–75.
Middle ton, G.V., Hampton, M.A., 1973. Sed i ment grav ity flows:
me chan ics of flow and de po si tion. In: Turbidites and Deep-wa - ter Sed i men ta tion (eds. G.V. Middle ton and A.H. Bouma): 1–38.
Pa cific Sec tion SEPM, Los An geles.
Migaszewski, Z., Ha³as, S., Durakiewicz, T., 1996. The age and or i gin of the cal cite min er al iza tion in the Holy Cross Mts based on lith o logic-petrographic and iso to pic ev i dence (Cen tral Po - land) (in Pol ish with Eng lish sum mary). Przegl¹d Geologiczny, 43: 275–281.
Mitchum, R.M., Vail, P.R., Sangree, J.R., 1977. Seis mic stra tig ra - phy and global changes of sea level, Part 6: Strati graphic In ter - pre ta tion of Seis mic Re flec tion Pat tern. AAPG Mem oir, 26:
117–133.
Nemec, W., 2009. What is a hyperconcentrated flow? In: Ab stract book, 27th An nual Meet ing of the In ter na tional As so ci a tion of An a logue GPR study of the Perm ian fanglomerates from Zygmuntówka Quarry near Chêciny... 765
Sedimentologists (eds. V. Pascucci and S. Andreucci), Alghero, Sep tem ber 20–23, 2009.
Nemec, W., Postma, G., 1993. Qua ter nary al lu vial fans in south - west ern Crete: sed i men ta tion pro cesses and geomorphic evo - lu tion. IAS Spe cial Pub li ca tion, 17: 235–276.
Nemec, W., Steel, R.J., 1984. Al lu vial and coastal con glom er ates:
Their sig nif i cant fea tures and some com ments on grav elly mass flow de pos its. Ca na dian So ci ety of Pe tro leum Ge ol o gists Mem - oir, 10: 1–31.
Pokorski, J., 1988. Rotliegendes stra tig ra phy in the north-west ern Po land. Bul le tin of Pol ish Acad emy of Sci ences, Earth Sci - ences, 36: 99–108.
Ronayne, M.J., Gorelick, S.M., Caers, J. 2008. Iden ti fy ing dis crete geo logic struc tures that pro duce anom a lous hy drau lic re - sponse: an in verse mod el ing ap proach. Wa ter Re sources Re - search, 44: W08426.
Simons, E.V., Rich ard son, D.B., 1966. Re sis tance to flow in al lu - vial chan nels. US Geo log i cal Sur vey Pro fes sional Pa per, 422:
1–61.
Straubhaar, J., Renard, P., Mariethoz, G., Froidevaux, R., Besson, O., 2010. An im proved par al lel mul ti ple-point al go rithm us ing a list ap proach. Math e mat i cal Geosciences, 43: 305–328.
Strebelle, S., 2000. Se quen tial sim u la tion draw ing struc tures from train ing im ages. Un pub lished Ph.D. thesis, Stan ford Uni ver sity.
Strebelle, S., 2002. Con di tional sim u la tion of com plex geo log i cal struc tures us ing mul ti ple-point sta tis tics. Math e mat i cal Ge ol ogy 34: 1–21.
Tay lor, G., Eggleton, R.A., 2001. Regolith ge ol ogy and geo mor - phol ogy. John Wiley and Sons Chichester, UK.
Thiry, M., Schmitt, J.M., Si mon-Coinçon, R., 1999. Prob lems, prog ress and fu ture re search con cern ing palaeoweathering and palaeosurfaces. Spe cial Pub li ca tions of In ter na tional As so ci a - tion of Sedimentologists, 27: 3–17.
Zbroja, S., Kuleta, M., Migaszewski, Z.M., 1998. New data on con - glom er ates of quarry “Zygmuntówka” in the Holy Cross Mts (in Pol ish with Eng lish sum mary). Biuletyn Pañstwowego Instytutu Geologicznego, 379: 43–59.
